Arbor, developed by Anandb71, maps code dependencies and predicts change impact across complex repositories, acting as a local MCP server for AI coding agents. The tool builds a semantic dependency graph to identify downstream effects and quantify blast radius, providing graph-backed symbol resolution, incremental updates, and Git-aware pull request analysis. The tool targets software engineers and technical architects working on interconnected codebases who use AI-assisted workflows and need earlier detection of regression risk.

What tasks can you actually use it for?

Arbor helps teams estimate which modules a code change will affect, directing review effort toward the highest-risk areas. It supplies AI coding agents with structured repository context via an MCP endpoint, so automated suggestions can reference graph links instead of plain text matches. Typical uses include triaging large pull requests, prioritizing files for manual review, and improving the precision of agent-generated patches.

How accurate are its impact predictions in practice?

Prediction quality depends on graph completeness and language coverage. The tool emphasizes semantic symbol resolution, which improves cross-file mapping in supported languages such as Rust, Python, and JavaScript/TypeScript. Because the dependency graph updates incrementally as edits occur, the analysis stays closer to the current repository state, which reduces stale inferences; accuracy declines when the project uses languages outside the stated support set.

Does it fit existing developer workflows and privacy needs?

Local-first deployment keeps source code on-device while exposing structured context to AI agents over MCP, aligning with teams that require on-premise processing. The server runs on macOS, Linux, and Windows and installs via Cargo, Homebrew, Scoop, or Docker. Integration requires an MCP-compatible client such as Claude Desktop, and the tool integrates with Git workflows to surface higher-risk merges during pull request review.

Who should adopt it and how to treat its outputs

Arbor is a practical choice for engineering teams that run MCP-capable assistants and manage tightly coupled codebases, because it supplies structured, local context to agents while keeping source code on-device. Projects that rely on languages outside its primary support set should validate mapping fidelity before depending on impact estimates. Treat Arbor's analysis as an architectural lens that augments, not replaces, human review.